The Design And Implementation Of Bulb Head Detection System Based On DM6446

On the product line of incandescent lamp in a lighting company, manual methods was used to test the quality of bulb head and eliminate substandard products because of unsuccessful welding. However, manual inspection not only cost much and was inefficient, but also couldn’t guarantee the accuracy. The purpose of this paper was to apply image processing technology combined with machine vision and embedded technology to develop an automatic detection device of bulb head which could remove unqualified products efficiently and accurately.The design of bulb head quality detection system was based on Texas Instruments ’s TMS320DM6446 with dual-core(DSP + ARM) and high performance, and this hardware system was constructed by using the spherical light source, bracket, camera, display and other detection device. Corresponding to the DSP and ARM core of main processing chip, software development was divided into the development of application running in the ARM core and development of image detection algorithm of light bulb head running in the DSP core.In the initial stage of the ARM side software development, the functions of detection system were analyzed attentively and carefully. Then, multi-threading architecture was used to implement the four parts of the ARM application. The Main function carried on the initialization of the whole system and derived three threads, which were Capture, Process and Display thread to deal with the video frames separately. The design job of Capture thread included the coding of camera based on V4L2 and the programming of the main loop. The design of Process thread was to accomplish the detection algorithm call based on Codec Engine mechanism and successfully exchange data with other two threads. As for Display thread, it was based on Frame Buffer and double buffer display technology. In addition, the paper also explored the synchronous initialization, working process and data interaction of three threads.According to the substandard conditions from the most frequent to normal, algorithm was designed to detect the products step by step. In the pre-processing stage of the algorithm, Hough Transform of circle was used to obtain two important circles and separate the bulb head image into annular black insulation region and center plate welding region. For annular black insulation region, algorithm was focused on the wire leakage. As for center plate welding region, algorithm was used to detect whether there existed hole or large-area without welding. Image processing algorithm was designed, developed and implemented in MATLAB software.